Development of automated UAV LiDAR based blueprint in oil palm replanting on terraces

Abstract Malaysia, which is the second largest palm oil producing country in the world, has faced the challenge of labor shortage to sustain palm oil production. Measuring replanting locations for palm seedlings manually in the field is a labor-intensive task, especially in sloping areas with terraces. However, algorithms or freely available software is limited. Therefore, this study aimed to develop programs to automate the generation of replanting points for palm seedlings, particularly on terraces in sloping areas using Python programming. Terraces were detected using a deep learning approach from a Curvature and Slope (CS) image generated from UAV LiDAR-based DEM. A series of processing steps were applied to remove error terrace lines and to refine them. On refined terraces, planting points were successfully generated automatically. The generated planting points were compared with the previous palm points. The results indicated that areas around the hilltop or foot of the hill, where the CS image was less distinct, tended to fail in generating points. Additionally, short terrace lines were unintentionally eliminated during processing. Although the programs need improvement, the proposed method successfully generated points across the entire study site, which is expected to reduce labor costs greatly. Further improvement of the program, additional training for the deep learning model, and trials of the proposed method at different sites are needed to enhance its robustness and applicability.